agent-harness

Use when the user says "/deep-review", "deep review", "thorough review", or wants the deepest possible code review before pushing a branch.

Use when the user says "/harness-update", "update the harness", "upgrade agent-harness", "pull the latest harness", or after seeing an UPGRADE_AVAILABLE notice. Pulls latest harness files into this project, preserves project-specific configuration and local-only skills, and walks the user through any conflicts.

Use when the user wants to audit or understand the agent harness itself — its hooks, skills, agents, or commands — or asks "what's in this harness", "how does this harness work", or "harness overview".

Use when the user has a one-off task that slots into the current sprint and needs a quick written plan — lighter than /plan-sprint. Reads existing plans for context, uses plan mode, writes a focused plan file.

Use when the user says "/build", hands off a sprint-plan document path, or asks to execute a written plan end-to-end. Branches, implements, tests, verifies in browser, commits incrementally, and prepares the PR. Auto-fires /lg-* skills when plan steps involve LangGraph/LangChain agent work.

Use when the user says "review migrations", "db review", "check schema changes", "safe to migrate?", or before deploying schema changes to production.

Use when encountering a bug, test failure, or unexpected behavior — runs staged investigation before implementing any fix.

Use when the user describes work that spans multiple files/systems, involves architectural decisions, would exceed 9 complexity pts or span multiple sprints, or says "deep plan", "plan this feature", or "let's plan [complex thing]". Produces folder-based plans with an entry-point doc plus individually executable sub-plans.

Use when the user asks for a demo script, customer story, or end-of-week demo for the current week's goals. Generates a 5-minute persona-told script scoped to what can ship in a week.

Use when the user says "start the dev server", "restart the server", "check the server logs", "what's the server error", or any dev-server lifecycle phrasing.

Use when the user says "verify", "check it", "test it", "e2e", or after completing UI/API changes that need browser-level confirmation before /ship.

Use when the user says "/grade-codebase", "grade this codebase", "how agent-friendly is this repo", "score the codebase", or wants a longitudinal measure of how well an LLM coding agent can work in this codebase.

Use when the user reports a production problem — "users can't log in", "500 errors", "the site is down", "something broke", "getting errors on [page]", "incident", or "production issue". Drives structured triage → diagnosis → remediation.

Use when the user says "/learn", "/learn <description>", "remember this", "save this lesson", or "add this to learnings" — captures corrections and surprising approvals from the current session into durable memory.

Use when the user says "add HITL to my agent", "make this graph durable", "wire streaming", "add a sub-agent", "attach a checkpointer", or any "add X to my LangGraph" phrasing. Adds capabilities to an existing LangGraph agent — HITL, persistence, streaming, sub-agents, custom tools, middleware, or BaseStore.

Use when the user asks "what's the right way to do X in LangGraph", "how do streaming/checkpointers/HITL work", or any general LangGraph reference question. Other lg-* skills load this for shared context (mental model, v1 API surface, footguns, deprecated patterns, JS/TS specifics, production checklist).

Use when the user says "design an agent", "I'm building an agent that does X", "help me architect this LangGraph", or before any agent implementation work. Asks structured questions, picks the right pattern (createAgent vs raw StateGraph vs Deep Agent vs supervisor/swarm), produces a design doc.

Use when the user says "add evals", "write a regression test for my agent", "set up LangSmith evals", "check the agent's trajectory", or "I broke something — write the test first". Sets up evals for a LangGraph/LangChain agent in either local-only or LangSmith-backed mode (trajectory, final-answer, smoke/hallucination, custom rule-based and LLM-as-judge).

Use when the user says "review my agent", "is this LangGraph code current", "audit this for footguns", "help me migrate from v0", or "find the bug in my graph". Reviews LangChain/LangGraph code for v1 best practices, deprecated patterns, and known footguns; doubles as the migration scout for v0→v1 upgrades.

Use when the user says "scaffold an agent", "build me a LangGraph agent that does X", "create a Deep Agent", "start a new agent", or hands off from /lg-design with a design path. Generates runnable TypeScript using LangChain v1 / LangGraph v1 patterns — createAgent (or raw StateGraph / Deep Agent) + tools + checkpointer + optional LangSmith tracing + streaming wiring.

Use when the user says "brainstorm this", "I have an idea", "help me think through this", "office hours", "is this worth building", or describes a new product idea before any code is written. YC-style strategic conversation in two modes (Startup diagnostic / Builder generative); saves a design doc.

Use when the user starts a new week or says "/plan-sprint", "plan the week", or "turn goals into projects". Breaks the current week's goals into concrete, executable projects with full implementation plans (file footprints, test criteria, E2E verification) under docs/plans/YYYY-wNN/sprint-plans/.

Use when the user says "pre-deploy", "ready to deploy?", "deploy check", "go/no-go", or before pushing to a production branch — runs the full pre-deployment quality gate.

Use when the user says "security review", "audit security", "check for vulnerabilities", or "is this secure?", or before any /ship to a production branch. Also fire proactively when changes touch auth, credentials, raw SQL, API routes, or external integrations.

Use when the user says "check the UI", "verify changes", "does it look right", or after completing UI work — quick browser spot-check (not a deploy gate; that is /e2e-verify).

Use when the user says "/ship", "ship it", "let's ship", or "push this up" and the branch is ready to leave the workstation — runs the full shipping pipeline (tests, lint, e2e, commit, push, PR).

Use when authoring or upgrading a rigid skill — runs subagent baselines under pressure to harvest real rationalizations before writing the Rationalization Table. Triggers on "baseline this skill", "TDD this skill", "add rationalizations to /<skill>", or before any new rigid skill ships.

Use when the user says "/skill-eval", "eval the skills", "run skill evals", "check /<skill> still works", or after editing a rigid skill's body — orchestrates the Phase 2 execution layer for `eval.yaml` files.

Use when the user says "/sync", "pull main", "sync with main", or "get latest". Switches to main and pulls the latest from remote.

Use when implementing any feature or bugfix, before writing implementation code, or when the user says "write tests first", "TDD", or "test-drive this".

Use when the user starts a new task and you need to align with weekly priorities, asks "what should I work on", says "/weekly-goals", or "check this week's goals". Loads and references the current week's goals; flag drift and suggest the next thing.

Use when spinning up builder agents, working on multiple isolated tasks simultaneously, or the user says "create a worktree", "isolated branch", or "parallel workspace".

Use when authoring a new skill, editing an existing skill, upgrading a flexible skill to rigid, or the user says "/write-skill", "create a skill", "new skill for X", or "make this a skill". Fires before any new `SKILL.md` is committed.